Mastering Midjourney V7: A Beginner's Guide to Prompt Engineering

Ready to take your Midjourney skills to the next level? This guide breaks down the essentials of prompt engineering in Midjourney V7. Learn how to craft effective prompts and troubleshoot common issues to bring your creative visions to life. Whether you're a beginner or looking to refine your skills, this article will equip you with the knowledge to create stunning AI-generated art.

Getting Started with Midjourney Prompting

Before diving in, it's important to understand how to interact with Midjourney and ask questions. The best way to learn is by doing, so let's get started with the basics of prompting and how to get help when you're stuck. Remember, there's a community of users ready to help if you get stuck. If you are ready to seriously dive into Midjourney, you may need to look at automation tools. The Midjourney Automation Suite from TitanXT can help automate and enhance the reader's Midjourney experience.

A Quick Warm-Up: Engaging with the Community

Before we get into the details, let's make sure everyone knows how to participate and ask questions. Feel free to drop a "Hi" or your favorite color in the chat to test things out and ensure you're ready to interact. We want to make sure you all know how to participate before we get into the good stuff. The Midjourney community is an amazing resource for tips and tricks, so don't be afraid to ask questions and share your creations.

The Addictive Nature of Midjourney

Have you ever felt like Midjourney is easy to binge? It's like binge-watching your own imagination! It can be pretty addicting. Did you know that Midjourney can literally change your brain? Neuroplasticity describes how your brain creates new physical pathways. If you immerse yourself in Midjourney, your brain may very well get better at noticing visual and artistic details everywhere because you are regularly stimulating your visual cortex. You might find that you have a heightened awareness that makes you more sensitive to colors, textures, shapes, and beauty in everyday life.

You might even look out a window or through a dashboard and suddenly be aware that the frame you usually take for granted is now framing a composition. Things might seem more artful, and you might ask yourself, "How would I prompt that?" or "That would make a good prompt."

Understanding Midjourney's Process

Let's start with a quick breakdown of how Midjourney works, so you understand why some prompts perform better than others. We'll go through some troubleshooting steps to consider when your prompt isn't working. We'll talk about controlling the whole canvas. If you don't control it with your prompt, Midjourney is just going to make stuff up. We want to learn some methods for trying to extend our reach to all parts of the canvas. We'll also talk a little bit about optimizing a prompt and using words that Midjourney understands.

By the end of this session, you'll have a troubleshooting framework. The fact is, what you want to do is, when your prompt isn't working, stop and consider a couple of things. Then, make a couple of changes in your prompt to see if you can get a little more success out of it.

The Bottom Line: Why Prompts Fail

Why do prompts sometimes not perform as preferred? First, if you don't control something in your prompt, Midjourney will fill in the gaps using its best guess. That means you need to make an attempt to control all parts of the canvas. All the details that you think are important, you should probably address with your prompt. We call that anchoring the details or pinning the details. Keep it on the canvas. When we say canvas, we mean the area of the image. We sometimes say no lazy prompting in order to help us remember that we actually have to make an effort to control all the parts of the canvas that we care about.

Each prompt is a timed job on Midjourney's servers. If your request takes too long to process, the AI will start blending and dropping details, or just getting weird, which we call incoherence. That's why there are some tricks for optimizing the prompt. Watch out for chaotic tokens, like conversational instructions that are not in conversational mode. Keep conversational instructions, like "make this a happy cat and put a crown on it," for conversational mode only.

You don't want prompts that read like a novel. Grammar and punctuation in full sentences matter because of how tokenization works. Midjourney will look at the structure of your prompt. So, a prompt like "sunset crime eagle city cyberpunk" is a jumbled mess. Midjourney is going to try to guess how those words relate. You're probably going to get some sort of eagle flying against the sunset above a city with neon aspects to it. But if you're trying to get something specific, you can't just put lists of words. You're going to have to add structure, like "a cyberpunk eagle perched on a rooftop, neon city lights glowing at sunset." This is clarity and specificity. To further help you, the Midjourney Automation Suite from TitanXT provides tools for refining and structuring your prompts effectively, ensuring clearer communication with Midjourney's AI.

How Midjourney Processes Your Input

Midjourney uses a process called diffusion to create images from your prompts. When Midjourney was trained, it learned how certain pixel patterns correspond to words. It looked at millions of images and matched those pixel patterns to text from alt text and nearby text, often captions to a picture. Over time, it built rules about the correspondence between the pixel patterns in the files and the words correlated to the files.

Now, when you type a prompt, Midjourney follows those rules to refine an image step by step. It passes over its starting point in waves, making billions of tiny adjustments to the pixels until those changes become so small that the model decides it's refined as it can get. That process is called denoising. The reason why this is called denoising makes more sense if we talk more about that starting point, which is called a seed. A seed is the random visual noise that Midjourney is denoising into an image. Midjourney never starts from a blank canvas. It's either starting from this random visual noise, the seed, or it's starting from an image it's already made.

If you use the same seed with the same prompt and settings, you'll theoretically get a copy of the image because it's starting from the same place. But there's a catch: Midjourney assigns your job to a random GPU every time you generate an image. That means your seed value isn't locked and isn't useful for maintaining consistency long term. The only value for the seed is probably in testing, because then we're sending our jobs fast enough that we might hit the same GPU and we might get the same seed.

Ideally, Midjourney is refining the seed in waves until it sort of plateaus, meaning there are no more meaningful changes happening. But in reality, chances are it just ran out of processing time based on the job you sent to the server. That's why you want to worry a little bit about prompt optimization.

Controlling the Canvas: Subject, Background, and Style

If it's not controlled with the prompt, Midjourney makes it up. But Midjourney isn't like a wildly imaginative avant-garde artist. It works from archetypes and stereotypes, using the most stereotypical version of whatever you asked for. Its imagination is only interesting when the juxtaposition of two incredibly normal things is interesting to us. If you don't control every part of a canvas, Midjourney doesn't get creative; it plays it safe. It's going to fill in the gaps with whatever it's seen the most in its training data.

Prompt Ordering and Structure

In V7, it's not immediately apparent in most prompts that prompt ordering has an effect. Older models of Midjourney are obsessed with prompt ordering, the sequence of words, what comes first, because they're used to this slope of influence that is greater at the front of the prompt. But in version six and certainly in version seven, that almost matters not at all because Midjourney has enough time and efficiency to view the whole prompt with equal attention. We worry about the order in which we put things basically to create a checklist for our human brains so that we don't forget to control all parts of the prompt.

Your prompts are going to need three things: how should it look (the style), what's in the image (the subject), and where is it and what's the situation (the context or background). If you miss one, Midjourney just shrugs and fills in the blanks with the most aggressively predictable thing possible. If you say a wizard, you're going to get some version of movie Gandalf. But if you do fill in all three things, you're not just waiting to see what Midjourney gives you; you're going to be telling it exactly what to create. A tool like the Midjourney Automation Suite from TitanXT can further refine your prompt structure, helping you to ensure that no detail is overlooked and that Midjourney understands your creative direction perfectly.

For example, a flat cartoon depicting an orange sailboat on a teal sea at night has the subject, the background, and the style. It doesn't need to be a sentence like this; it could be an entire paragraph. But it needs to have the subject, the background in which the subject is contextualized, and the style in which the whole image is depicted. That'll help us get some control.

Leveraging Compound Subjects

One of the cool ways that you can get multiple subjects into your prompt without running out of time is to take clever advantage of compound subjects. Instead of describing a bunch of ships, you can just say an armada. Instead of saying a man, a woman, and two children, you could use the word family. Instead of saying a bunch of people in dark clothes stand around holding flowers and looking sad, you can just say a funeral.

Those are compound subjects. They're words that automatically give you more on the canvas without eating up precious processing time. A family, a team, a committee—it's kind of like a cheat code for more visual storytelling. So be clever in your use of archetypes when you're prompting.

Optimizing Your Prompts: Archetypes and Visual Language

Let's talk about optimizing the prompt, focusing on archetypes and how to use words that Midjourney understands. This is what you do in troubleshooting when you notice you're losing details or seeing blending. At the very least, there are a few ways to address this problem. A simple way to address this problem is to take advantage of archetypes.

Understanding and Using Archetypes

An archetype is the dominant representation of the thing in Midjourney's data set and rules. You can either describe the thing yourself or invoke the archetype and let Midjourney take care of it with stereotypical details. Learning when to describe versus invoke is an important part of prompting. Use the invoke method to make your prompt more efficient. Using the word lumberjack and letting Midjourney supply all the default archetypal details uses less processing time than describing the lumberjack yourself.

At the same time, you want to learn to avoid them and use the describe method to control undesirable outcomes in the images you create. This way of thinking of recognizing archetypes is asking you to look at the world and see it in terms of its patterns, the same way Midjourney looks at image files and sees pixel patterns. So prep your brain and be ready to meet the AI halfway.

Avoiding Chaotic Tokens

If the prompt contains chaotic tokens, you're going to lose control of the canvas. Chaotic tokens are words and phrases that Midjourney doesn't actually know how to translate into visuals. Things like "make sure the lighting is dramatic" don't work in traditional prompting mode. Also, jargon like "f1.8 aperture, 16-bit linear" won't work. Abstract concepts are tricky too. Midjourney doesn't know what regret looks like. If you don't define that sort of thing in the prompt, Midjourney is just going to shrug and throw something archetypal at you.

The fix is to use words that Midjourney actually understands. Instead of saying "a sorrowful night longing for home," try "a solitary knight wearing battered armor standing on a foggy battlefield in the dawn light." Midjourney has concrete things to work with. You want concrete visual things, a specific pose, a specific setting, a specific atmosphere. You're not hoping it understands sorrowful; you're giving it visual details that communicate that emotion. If your prompt sounds like it needs to be read in a dramatic monologue, stop and reconsider. Remove conversation that belongs in conversational mode, use real sentences, and use dense visual language.

Conclusion

Mastering prompt engineering in Midjourney V7 involves understanding how the AI processes your requests, controlling the canvas with specific details, and optimizing your prompts for efficiency. By focusing on clear, visual language and leveraging archetypes, you can create stunning and coherent images that match your creative vision. Remember, the key is to experiment, iterate, and continuously refine your approach to achieve the best results. Ready to take your Midjourney skills to the next level? Don't forget to explore tools like the Midjourney Automation Suite from TitanXT for enhanced control and efficiency.